Testing Models at Middlesbrough.
The specification for the Sydney Harbour Bridge required the successful contractors to provide a testing machine of 1250 tons (1270 tonnes) capacity to verify the design of bridge members. Since no machine could be created strong enough to test full size members, models, at as large a scale as possible which the machine could break, were used. In the event, even at a scale of 1 in 6 the machine could not break the models of the bottom chord at its greatest dimensions.
The machine, of nominal 2,800,000 pounds capacity, was designed and manufactured by W & T Avery Limited of Birmingham and installed at the Britannia Works of Dorman Long & Co at Middlesbrough in July 1926 in a three-storey brick building eventually known as the Britannia Test House.
It was at the time the largest Universal Testing Machine in the country and could handle specimens in both tension and compression up to 50 feet (15m) in length and as large as 3 ft. 9 in. (1150mm) square in cross section. Bend test specimens could be accommodated up to 20 ft. (6m). It was hydraulically powered, and the load was weighed off on a series of multiple levers counterpoised by means of a travelling jockey weight upon a graduated steelyard with great accuracy throughout the load scale, including small loads.
For the Sydney Harbour Bridge the typical regime was for three simpler models, without details such as intermediate splices in long heavy members, to be fabricated and tested to obtain consistency in the results and a single model with more details manufactured to verify that the details did not significantly change the result. In the bridge proper the heavier members of the lower chord near the bearings were erected in two or three sections, limited by the capacity of the creeper crane, but in the models, only a few per cent of the weight, there was no such limitation, so they were fabricated whole. In the one, more accurate model, the splices and the joints with their cover plates, as well the splice plates and gussets at the panel points were modelled as well as they could be. At a 1:6 scale some of the component steel sections which went into the assembly were too small and thin to be rolled. All sections used in the tests were rolled in silicon steel, identical to that used in the bridge. The models were riveted with 3/16 inch rivets or larger as appropriate.
Because of this limitation of the component sections at 1:6, further scale models were made at a larger scale of one-half of the chord section utilising material of more normal thickness. These were within the capacity of the machine to break. Half of the end member of the bottom chord (28-26) was made at 1:4.8 and the chord from the sixth panel (18-16) at a scale of 1:4.
In the ordinary models there was no attempt to reproduce the end-connection material. The material of the member alone was carried through and carefully square ended against the platens of the testing machine. The fourth model of each representative component, top chord, bottom chord, post or diagonal, was prepared in which the details were reproduced as closely as possible. Post 20-21 was chosen to study the effect of intermediate splices upon the strength of compression web members in the truss and so three plain models and three detailed models were tested.
The only tension member to be tested was diagonal 9-6. Although like all members of the truss, the diagonals are built in two halves, they act practically independently, and so only one was tested. For the testing exercise this had the advantage that the model could be built at as large a scale as 1:2 and still be within the capacity of the machine to break.
In the case of one model of each of the chord members tested, an attempt was made to reproduce in the test the bending stress produced in the actual bridge member due to its own weight. Loads were added to these models, which were tested in a horizontal position, to produce the same calculated stress intensity as that in the actual member, upon the assumption that its ends were freely supported.
Records of the distortion of the models were obtained from strain-gauges fixed near the outsides of the flanges of the members, utilising generally a gauge-length of approximately 10 feet. Readings of axial distortion to the nearest .0001 inch were taken direct by means of four " Ames " dials, and load-strain curves plotted up to the point at which the member ceased to be elastic. These results are shown in the adjacent graphs, figures 15 and 16. Records of the lateral distortions of the members were also taken, as well as numerous records of stresses in lacing bars and tie-plates, measured on short gauge-lengths.
Generally, what was learned from the tests was that the designs were adequate. No revisions in the design were made as a result of the tests.
Freeman stated in his paper:
A series of models were made and tested to destruction in the contractors' works at Middlesbrough, but the information gained, although of considerable interest, did not justify the alteration in design of any member tested.[1]
One important number which was defined was the effective Young’s Modulus of the members. Tests were made on eight small test-pieces, cut from silicon steel plates and angles, to determine the modulus of elasticity of the steel itself. The values obtained for these specimens varied from 30,400,000 to 31,600,000 lbs. per square inch, the average being 31,000,000 lbs. per square inch. The model member tests gave an apparent value of the modulus considerably lower than that of the steel as a material, due presumably to the effect of rivet-holes and slip between the component parts of the member. It was noticeable that the posts, which were of a simple section with fewer rivets, gave a higher modulus than the chord models, while the model of diagonals tested in tension had the lowest modulus of all. Also, the modulus was least for the first time the member was stressed, and increased progressively with repeated loading up to a maximum of about 30,000,000 lbs. per square inch.
A workable value of the modulus of elasticity was required for the calculations of deformation of the half arches as they reached out from the shore; the stretch of the cables as the load increased; the required set back of the end posts to allow space at the crown of the arch for closure; and the force required to be introduced by jacks to stress the top chord into its final designed state.
It would seem most likely that 39 models were tested, though Bradfield in his 1934 paper to the Institution of Civil Engineers lists 41. John Freeman Pain’s and Gilbert Roberts’ paper in the same series lists only 39. Bradfield shows three models of the unbreakable lower chord (B), but omits two models of half the chord 20-22 (R) which were breakable. He also shows three models of the heaviest top cord 3-5 (G) with all its splice and end connections included. This would seem to be unusual as in no other case was more than one detailed model tested.[2]
Other important components of the bridge erection gear to be tested were the tie-back cables which were designed to carry a maximum load of 128 tons. Numerous samples were tested to destruction, with none failing below 350 tons. The design of the end sockets for the cables was tested and although near failure there was evidence of distress in the socket, the failure was always in the cable. The modulus of elasticity of the cables was established as 8,500 tons per square inch for first loading rising to 9,000 tons per square inch.
Also studied was creep in the cables with prolonged load, as would be imposed on the bridge. For a load of 100 tons, about that applied in Sydney, they extended approximately 0.002 inch per foot in length in 40 days, with one half of this strain occurring in the first ten days. After 40 days extension practically ceased.
Critical tools in the erection procedure were the 850-ton jacks which were used to stress the top chord. The force which was required would be calculated and needed to be applied accurately. The eight jacks were calibrated using the Avery machine, with dedicated pressure gauges assigned to each.
While almost all of the Dorman Long steelworks and fabrication shops at Middlesbrough has been closed and demolished, the testing machine and its building survives in use, now operated by Durham Lifting.[3]
The machine has been upgraded to be able to test to 3,000 tons.
- ↑ ICE Journal 1934 Sydney Harbour Bridge: Design of the Structure and Foundations, P160
- ↑ The Dorman Long book about their achievement in building the bridge published in 1932 states ‘…nearly forty scale models…’
- ↑ https://www.durhamlifting.co.uk/
